Exemplar for +Inf bucket adds bucket with wrong value

[balintzs]

When writing histogram metrics, and there is an exemplar for the +Inf bucket, then the +Inf bucket is added with the value of the previous bucket:
https://github.com/prometheus/client_golang/blob/main/prometheus/metric.go#L188

This is incorrect, as the cumulative count of the +Inf bucket should instead be added with the total count of the histogram datapoint.

The above behaviour results in invalid values being reported for the +Inf bucket.

[balintzs]

I'll open a PR shortly

[bwplotka]

Thanks! Approved fix with some small nit.

[balintzs]

Thanks for merging it, @bwplotka. When can we expect it to get released?

[bwplotka]

Can you just use commit sha for now? We will release soon, but don't know when (KubeCom + PromCon is coming)

[bboreham]

Note your link to the problem now shows the fixed code, because it references main.
You can get a fixed link by pressing 'Y' in the GitHub UI, e.g.:

client_golang/prometheus/metric.go

Lines 188 to 190 in 3faf3ba

	// The +Inf bucket should be explicitly added if there is an exemplar for it, similar to non-const histogram logic in https://github.com/prometheus/client_golang/blob/main/prometheus/histogram.go#L357-L365.
	b := &dto.Bucket{
	CumulativeCount: proto.Uint64(pb.Histogram.Bucket[len(pb.Histogram.GetBucket())-1].GetCumulativeCount()),